AppJuice Direct AI Workbench
AppJuice Direct AI Workbench lets you try out the AI capabilities available in AppJuice using your own configured providers. You can generate or rewrite text, analyze and create images, transcribe or generate speech, create music, generate short videos, run web searches, read public pages, and evaluate structured decisions. Results appear right in the app, and you can build a simple report to copy or export.
Listing generated with opencode-go/deepseek-v4.1-flashReviewed with deepseek/deepseek-v4-pro
Need the desktop app first? Download AppJuice for macOS, install it, then come back to install this app through AppJuice.
What you can do
- Generate, rewrite, summarize, or classify text with prompts and optional conversation history.
- Upload images and ask questions about them.
- Create images from prompts, with optional reference images for style guidance.
- Transcribe audio files or record your voice directly in the app.
- Turn text into spoken audio with voice and format options.
- Generate music from a prompt and optional lyrics.
- Create short videos from prompts, with optional reference images and resolution choices.
- Search the web for current information and read text from public pages.
- Evaluate structured decisions with yes/no, choice, and scored question presets.
AI and local access
- The app runs locally in AppJuice and uses the AI providers and keys you have already configured.
- Your provider credentials are never stored in the app; AI requests are handled securely through AppJuice.
- Uploaded files and recordings are used only for the selected AI task and are not shared beyond the provider you choose.
- Web search and page reading use AppJuice's configured web backends.
- Generated images, audio, and video can be previewed and downloaded.
Safety and privacy
- No hardcoded API keys, tokens, or private files were found in this release.
- The app does not browse your local file system or folders; it only handles files you explicitly upload or record.
- Results are kept temporarily while you work and are not sent anywhere except to the AI provider you selected for that test.
- No automatic background agent is enabled; every action runs only when you press Run.
Good to know
- Some AI capabilities depend on the providers and models you have configured in AppJuice; availability may vary.
- Edge TTS is available as a credential-free fallback for text-to-speech.
- Video generation may return a file immediately or a job you can poll; options like resolution and aspect ratio depend on the provider.
- Web search and web extract do not require a model selection; they use AppJuice's web settings.
- Structured decision presets use the AppJuice decision-making recommendation.
Version history
Latest - v1.12.2
This update adds a Decision Making test card for evaluating yes/no, multiple-choice, and scored questions using your configured AI provider.
- Added a Decision Making capability with editable example questions.
- Included examples for support triage, message signals, request routing, and bug severity.
- Results now show probabilities, confidence, and score breakdowns in the report.
- Updated the app description and plans to include decisions.
Previous - v1.11.7
This update improves the workbench with new web tools, a cleaner report layout, dark theme support, and broader provider and video options.
- Added Web Search and Web Extract test cards.
- Renamed the app from Direct AI Helper Demo to Direct AI Workbench.
- Moved the report into its own sidebar panel with a run count.
- Added light/dark theme and AppJuice text-size support.
- Expanded video generation options including reference images, aspect ratio, resolution, polling, and download controls.
- Updated text-to-speech with speaking instructions and a longer provider list.
- Added license, README, and favicon, and refreshed the app icon.
Previous - v1.7.4
This is the first public release of the AppJuice Direct AI Helper Demo.